Seeds of commercially acquired carrot (Daucus carota L.) cultivar Nantaise were surface sterilized in a solution of 95% ethanol for 1 min and washed once with sterile water for 3 min. Then, the seeds were incubated under agitation in a solution of sodium hypochlorite (2.62% v/v) for 45 min, washed three times with sterile water and finally dried on sterile absorbent paper. The sterile seeds were deposited in sterile flasks with solid MS medium (Murashige and Skoog, 1962) supplemented with 0.44% vitamins, 2% sucrose, 0.01% myo-inositol, 0.7% Agar and pH adjusted to 5.8. The seeds were kept in a growth chamber 3 weeks with a 16 h long day photoperiod illuminated with cool-white fluorescent light (115 μmol m^-2 s^-1) at 22°C. Hypocotyls and stems of 3-week-old in vitro wild-type carrot plantlets were utilized in Agrobacterium-mediated transformation experiments. Transformed carrots were transferred to pots (20 × 10) and cultivated in the greenhouse, as described above.

Binary vectors for constitutive expression of Arabidopsis genes encoding DXS or DXR proteins fused to GFP were previously reported (Pulido et al., 2013; Perello et al., 2016). Agrobacterium tumefaciens (strain GV3101) cells were transformed with these vectors and used for D. carota transformation following the protocol described by (Chen and Punja, 2002). Briefly, hypocotyl segments of 3 weeks-old seedlings were co-cultivated with Agrobacterium carrying the vector of interest, and placed on solidified MS media (4.4 g/L MS salts, 20 g/L sucrose and 0.7% agar) in darkness. After 2 days, the explants were transferred to solid MS medium containing 1 mg/L 2.4D for somatic embryogenesis induction and supplemented with 0.5 mg/L Basta^® and 300 mg/L cefotaxime. After 4 weeks in darkness, the explants were placed on solidified MS medium containing 0.5 mg/L 2.4D, 1 mg/L Basta^® and 300 mg/L cefotaxime in photoperiod conditions (16 h light, 115 μmol/m²/s). Herbicide-resistant embryos were transferred to MS media in the absence of hormones to induce the development of shoots. After 6 months, transformed plantlets were transferred to soil in a temperature and photoperiod controlled greenhouse (16 h light, 115 μmol/m²/s). Different stages of the procedure are shown in Supplementary Figure S1. We obtained several lines per construct and PCR analyses to confirm the presence of the transgenes in the transgenic lines were performed with primers AtDXRF, AtDXSF, and eGFPR (Supplementary Figure S2; Supplementary Table S1). Transgenic lines developed normally and were visually undistinguishable from non-transgenic controls (Supplementary Figure S3). Four of the lines confirmed to contain the corresponding transgene by PCR (Supplementary Figures S2 and S3) were then picked for quantification of transcript and pigment levels. From those, three representative lines per construct were selected for more detailed analysis.

Photosynthetic pigments from leaves and roots of wild-type and transgenic plants transferred to soil and grown in the greenhouse for 6 months were extracted from 100 mg of fresh weight with 1 ml of hexane/acetone/ethanol (2:1:1 v/v) as described (Fuentes et al., 2012). The extract was dried with N₂. To quantify the concentration of chlorophyll a, chlorophyll b and total carotenoids present in the pigment extracts of leaves and roots of D. carota, the pigments extracted were resuspended in 2 mL of acetone and using a spectrophotometer the absorbance was measured at 750, 662, 645, and 470 nm in quartz cuvettes. Absorbance at 662, 645, and 470 nm is used to determine the concentration of chlorophyll a, chlorophyll b and total carotenoids, respectively. Further, the absorbance at 750 nm was measured to determine the turbidity of the sample because the turbid samples may result in underestimation of the concentration of the pigments concentration. With absorbance measurements, the concentration of chlorophyll a, chlorophyll b and total carotenoids determined by the equations described (Lichtenthaler and Buschmann, 2001). For α-carotene and β-carotene measurements, the pigments were separated by a HPLC using a RP-18 Lichrocart125-4 reverse phase column (Merck^®), utilizing a acetonitrile: methanol: isopropanol (85:10:5 v/v) mix as a mobile phase with a 1 ml/min flow rate at room temperature in isocratic conditions. The elution spectra of each maximum were obtained using a diode array detector. The carotenoids were identified according to their absorption spectra, retention time and comparison with specific pigment standards, which was corroborated by comparison with the Carotenoids Handbook (Britton, 1995; Britton et al., 2004). All operations were carried out in triplicate, on ice and dark conditions to avoid photodegradation, isomerization and structural changes of carotenoids.

A frozen powder of 100 mg of D. carota leaves from plants transferred to soil and grown in the greenhouse for 6 months was used for total RNA extraction using TRIzol^® reagent (Invitrogen) and following the manufacturer’s instructions. For cDNA synthesis, 2 μg of total DNA-free RNA was mixed with 1mM of oligodT primer and Impron II reverse transcriptase (Promega). The expression of the DXS and DXR transgenes was estimated by RT-PCR using primers qDXRF, qDXSF and qeGFP R (Supplementary Table S1). Quantitative RT-PCR (qRT) experiments were performed in a Stratagene Mx3000P thermocycler, using SYBR Green double strand DNA binding dye as described previously (Stange et al., 2008). Specific primers for carrot genes were designed targeting the 5′ UTR of PSY1 (AB032797) and PSY2 (DQ192187) and the coding sequence of the 18S gene, selected as the normalizer (Supplementary Table S1). Final data were obtained introducing fluorescence results in the equation described by Pfaffl (2001). Each qRT-PCR reaction was performed with three biological replicates and each sample was analyzed in duplicate (technical replicate). In all cases, the reaction specificities were tested with melting gradient dissociation curves and electrophoresis gels. To test for significant differences in gene expression, results were analyzed using the General Linear Models option in the statistical software package Graphpad Prism. The one and two tailed Student t-test (p < 0.05, confidence interval 95%), were used.

In order to determine the significance of the supply of MEP-derived metabolic precursors for the synthesis of carotenoids (and chlorophylls) in carrot, we generated transgenic plants expressing GFP-tagged versions of the Arabidopsis DXS/DXS1/CLA1 (At4g15560) or DXR (At5g62790) genes under the control of the constitutive 35S promoter (Supplementary Figures S1 and S2). The presence of the corresponding transgene in the generated S (35S:DXS) and R (35S:DXR) lines was confirmed by PCR (Supplementary Figure S2). After semi-quantitative RT-PCR analyses to estimate transgene expression levels in 6-month-old carrot plants, three representative lines of each construct were selected for further experiments (Figure 2). First, we aimed to quantify the accumulation of carotenoids in the root of transgenic plants and untransformed wild-type (WT) controls. As shown in Figure 3A, the constitutive expression of DXS produced a substantial increment in α-carotene (up to 3.6-fold) and β-carotene (up to 2.7-fold) levels in the storage root. As a consequence, total carotenoids were also significantly higher than in untransformed controls, reaching 2600 μg/g DW (dry weight) in line S29, the one also showing the highest levels of transgene expression (Figure 2). On the other hand, the constitutive expression of DXR only led to a slight increase in β-carotene in line R32 (the one with highest transgene expression levels) but no significant changes in the levels of α-carotene and total carotenoids in the root (Figure 3B). When calculating the mean of all transgenic lines together, the constitutive expression of the DXS gene resulted in an average increase of 114% of α-carotene, 75% of β-carotene, and 78% of total carotenoids with respect to WT plants, whereas DXR overexpression had no impact on root carotenoid levels (Figure 3). These results suggest that DXS, but not DXR, plays an important role in controlling the flow of metabolic precursors toward carotenoid synthesis in the storage root of D. carota plants.

We next asked whether the upregulation of DXS or DXR levels could have a different effect on photosynthetic tissues of carrot plants. To address this question, we quantified the levels of both carotenoids and chlorophylls in leaves of the selected transgenic lines. As shown in Figure 4A, the constitutive overexpression of DXS produced a significant raise in the concentration of total carotenoids. In particular, α- and β-carotene increased in leaves of all lines tested, reaching levels that were up to fourfold and twofold higher, respectively, than those found in untransformed controls. Levels of lutein, however, were only significantly higher in line S29 (Figure 4A). On the other hand, the overexpression of DXR produced modest but statistically significant increments in α-carotene, β-carotene and lutein in leaves of some of the transgenic lines tested, giving rise to a slight increment in total carotenoids in all transgenic lines (Figure 4B). On average, the expression of the Arabidopsis DXS gene in D. carota produced an increase of 91, 69, 30, and 80% in the concentration of α-carotene, β-carotene, lutein and total carotenoids, respectively, while increasing DXR levels only led to statistically significant increments in lutein (17%) and total carotenoids (19%; Figure 4).

Similar results were found for chlorophylls. Thus, the constitutive overexpression of DXS significantly increased the accumulation of chlorophylls in the leaves of all transgenic lines tested (Figure 5A). In particular, the levels of chlorophyll a in the best performing line (S29) were about twofold higher compared to the WT, whereas a 1.4-fold rise in chlorophyll b levels was detected in the same line (Figure 5A). The overexpression of DXR resulted in only minor changes in chlorophyll levels. While chlorophyll a did not change in any of the lines tested, modest but significant increases in the concentration of chlorophyll b and total chlorophylls were detected in some lines (Figure 5B). On average, increasing DXS activity led to 78% more chlorophyll a, 40% more chlorophyll b and 63% more total chlorophylls in carrot leaves, while increasing DXR activity did not change the levels of chlorophyll a but produced 38% more chlorophyll b and 21% more total chlorophylls compared to untransformed leaves (Figure 5).

While DXS has been shown to be the main rate-determining enzyme of the MEP pathway, PSY plays a similar role in the carotenoid pathway (Fraser et al., 2002; Wright et al., 2014). In fact, an increased DXS activity would only result in increased carotenoid biosynthesis if PSY activity is high enough to channel the extra supply of MEP-derived precursors to the carotenoid pathway. Consistent with the central role of these two enzymes in the control of the carotenoid pathway flux, a recent genome-scale analysis (Iorizzo et al., 2016) showed that carrot genes encoding DXS (DCAR_030576, named DXS1) and PSY (DCAR_023043 and DCAR_010057, respectively, named PSY1 and PSY2) were differentially expressed in storage roots from populations with different carotenoid levels, showing increased transcript levels in accessions with higher carotenoid levels (Supplementary Figure S4). By contrast, no changes were found for the only carrot gene encoding DXR (DCAR_026133). Because transgene-mediated alterations of the carrot carotenoid pathway have been associated with a concomitant alteration of the transcript levels of endogenous PSY1 and PSY2 genes (Moreno et al., 2013), we next analyzed whether the transcript abundance of these two PSY-encoding genes was changed in leaves of representative lines overexpressing the Arabidopsis DXS and DXR genes. As shown in Figure 6A, leaves from transgenic plants overexpressing DXS showed a dramatic increase in the relative abundance of PSY1 transcripts compared to WT controls. A lower increase was observed for PSY2 transcripts (Figure 6A). In general terms, a fairly good correlation was found between levels of exogenous DXS transcripts (Figure 2) and endogenous PSY-encoding gene expression (Figure 6A) in leaves. Unexpectedly, however, we observed that the constitutive overexpression of DXR did not result in higher but lower levels of endogenous PSY1 and PSY2 transcripts in leaves from lines R32 and R37 and no changes in line R33 (Figure 6B).